| The dissertation is aimed at advancing knowledge of the correlates of crime at small geographic units of analysis. I begin by detailing what motivates examining crime at small places, and show how aggregation creates confounds that limit causal inference. Local and spatial effects are confounded when using aggregate units, so when the researcher wishes to distinguish between these two types of effects it should guide what unit of analysis is chosen. To illustrate these differences, I generate simulations of what happens to effect estimates when you aggregate a micro level spatial effects model or presume a neighborhood effects model.;I provide further examples in case studies that examine local, spatial and contextual effects for bars, broken windows and crime using publicly available data from Washington, D.C. Using negative binomial regression models, I estimate that adding a bar to a street unit (street midpoints and intersections) increases the number of Part 1 crimes per year on the local street by around 1 on average, but increase the sum of crime on neighboring streets by 2. I also provide estimates of the selection of bars into criminogenic neighborhoods using a non-equivalent dependent variable design, and estimate the selection effect is 25 percent.;Using 311 calls for service as a proxy for physical disorder, I estimate their effects on crime using fixed effects for omitted neighborhood level variables. 311 calls for service have a consistent positive effect on crime using several different neighborhood boundaries, but the effects are very small. I also show the fixed effects model is not a good fit to the data, and that it potentially introduces artefacts at the boundaries of neighborhoods.;I end the dissertation by building a general model of crime including a variety of variables as well as non-linear terms to account for spatial trends. I show how this model corrects for poor predictions and spatial autocorrelation that was apparent in prior models. In these models prior estimates of the effect of bars on crime are slightly moderated, but the effects of 311 calls for service are still similar in size. |
